Chronic pancreatitis (CP) is a progressive and destructive inflammatory disorder of the pancreas. One of the most distressing features is pain, which occurs in ~90% of patients, about half of whom have constant pain, which is associated with increased hospitalization and lower quality of life. Since there are multiple pain etiologies and since no standardized method exists to assess pain in patients with CP, novel approaches are needed to better understand the mechanisms of pain in CP and how it can be appropriately treated. The North American Pancreatitis Study 2 (NAPS2) recruited, phenotyped, and obtained biospecimens from the largest US cohort of pancreatitis, and we have completed a genome-wide association study (GWAS) that includes 1,171 NAPS2 CP cases for whom detailed phenotypic pain data are available. We propose to analyze the NAPS2 data set and biospecimens (DNA, serum, pancreas tissue) to define genetic risks and potential mechanisms of constant pain, identify markers of inflammatory pain, and characterize clinical pain complexes that can guide patient management. We have already identified through the full and a nested GWAS 8 candidate genes for constant pain. Aim 1 will determine whether these variants are associated with functional changes in genes associated with the constant pain phenotype. Targeted SNP genotyping and gene expression studies will be conducted, including mRNA studies and immunohistochemical staining in human samples. Aim 2 will determine whether c-reactive protein (CRP) or cytokine biomarkers correlate with constant pain. Pain is an indicator of active inflammation, with pro-inflammatory cytokines increasing pain, and anti-inflammatory cytokines diminishing pain. We will test 500 NAPS2 samples for elevated CRP and 10 Th1/Th2 cytokines as biomarkers of inflammation. For positive controls we will include blood samples from 40 acute pancreatitis patients who remain hospitalized for > 4 days for pain or inflammation. Aim 3 will apply machine-learning approaches to test for correlation of genotype, biomarkers, and morphology (obstruction) with quantitative measures of pain pattern, severity, and character as well as SF12 v2 quality of life scores (mental and physical) while controlling for sex, smoking, and alcohol. We anticipate that our machine learning approaches will provide decision rules for the proper classification of pain according to etiology worthy of formal testing in clinical trials. In addition, use of machine learning is anticipated to provide insight into pain mechanism by optimally linking the symptoms signatures, biomarkers, imaging studies, and genetics to complex mechanisms that are seen in patients with painful CP. The goal of this study is to use existing NAPS2 data and biospecimens to construct a framework for future clinical studies that test the effectiveness of personalized pain management based on our machine-learning-predicted etiology rather than symptoms alone.